| Science |
AI in Medical Imaging Can Deliver Misleading Results
Artificial intelligence can enhance medical imaging by detecting patterns beyond human perception, but a new study reveals the technology may produce highly accurate yet misleading results by relying on unintended data cues.
AI in Medical Imaging Can Deliver Misleading Results. Photo: tasnimnews.com
A study highlights a challenge in using AI for medical imaging, known as "shortcut learning," where models exploit subtle and unrelated data patterns to make predictions.
Researchers analyzed over 25,000 knee X-rays and found that AI systems could "predict" implausible traits, such as whether patients abstained from eating refried beans or beer.
While these predictions lack medical relevance, the models demonstrated surprising accuracy by identifying unintended patterns in the data.
"While AI has the potential to transform medical imaging, we must be cautious," said Dr. Peter Schilling, the study's senior author, an orthopedic surgeon at Dartmouth Health's Dartmouth Hitchcock Medical Center.
"These models can see patterns humans cannot, but not all patterns they identify are meaningful or reliable," Schilling added.
The study found that AI algorithms often rely on confounding variables—such as differences in X-ray equipment or clinical site markers—rather than medically significant features.
Attempts to remove these biases were largely unsuccessful, as the models adapted by identifying other hidden patterns.
"This goes beyond bias from clues of race or gender," said Brandon Hill, a machine learning scientist at Dartmouth Hitchcock and co-author of the study.
"We found the algorithm could even learn to predict the year an X-ray was taken," Hill explained.
"When you prevent it from learning one of these elements, it will instead learn another it previously ignored. This danger can lead to some really dodgy claims, and researchers need to be aware of how readily this happens when using this technique."
The study emphasizes the importance of rigorous evaluation standards in AI-driven medical research.
Overreliance on standard algorithms without deeper scrutiny could lead to inaccurate clinical insights and flawed treatment decisions.
"The burden of proof just goes way up when it comes to using models for the discovery of new patterns in medicine," Hill said.
He added that part of the problem stems from human assumptions about AI.
"It is incredibly easy to fall into the trap of presuming that the model 'sees' the same way we do. In the end, it doesn't," Hill said.
"AI is almost like dealing with an alien intelligence," he continued.
"You want to say the model is 'cheating,' but that anthropomorphizes the technology. It learned a way to solve the task given to it, but not necessarily how a person would. It doesn't have logic or reasoning as we typically understand it." (Tasnim News Agency)
YOU MAY ALSO LIKE
